Higher Teichmüller Theory: Hitchin & Maximal Components¶
Job: construct the two principal families of higher Teichmüller spaces in depth — Hitchin components \(\mathrm{Hit}_n(\Sigma)\) for split real \(G\) (via principal \(\mathrm{SL}_2\), Higgs bundles, and the Hitchin section by holomorphic differentials, with Labourie's \(B\)-Anosov theorem) and maximal components for \(G\) of Hermitian type (via the Toledo invariant and the Milnor–Wood inequality, with the Burger–Iozzi–Wienhard discreteness theorem) — and record the non-abelian Hodge theory bridge (Corlette, Donaldson, Hitchin–Simpson) that identifies reductive representations with polystable Higgs bundles.
Core¶
Hitchin components¶
Let \(G=\mathrm{PSL}_n(\mathbb{R})\) (the prototypical split real \(G\)). The principal \(\mathrm{SL}_2\) is the irreducible \(\mathrm{PSL}_2(\mathbb{R})\hookrightarrow\mathrm{PSL}_n(\mathbb{R})\) acting on \(\mathrm{Sym}^{n-1}\mathbb{R}^2\cong\mathbb{R}^n\). The Fuchsian locus is the image of \(\mathcal{T}(\Sigma)\) under this embedding:
Theorem (Hitchin 1992). The connected component of \(\mathrm{Rep}(\pi_1\Sigma_g,\mathrm{PSL}_n(\mathbb{R}))\) containing the Fuchsian locus is a smooth manifold of dimension \((2g-2)(n^2-1)\), consisting of discrete and faithful representations. It is the Hitchin component \(\mathrm{Hit}_n(\Sigma)\). It is parametrized by the Hitchin section: a tuple of holomorphic differentials $$ (q_2,q_3,\dots,q_n),\qquad q_k\in H^0(\Sigma,K^{\otimes k}),\quad q_2\not\equiv 0, $$ where \(K\) is the canonical bundle of \(\Sigma\) (with the conformal structure varying over \(\mathcal{T}(\Sigma)\)). The Fuchsian locus is the slice \(\{q_3=\cdots=q_n=0\}\) with \(q_2\) the quadratic differential of the Fuchsian uniformization.
The parametrization is via the Hitchin fibration on the Higgs-bundle moduli space (below): the base of the fibration is \(\bigoplus_{k=2}^n H^0(K^{\otimes k})\), and the Hitchin section is a section whose value at \((q_2,\dots,q_n)\) is the Higgs bundle whose associated representation is the Hitchin representation. The dimension is \(\sum_{k=2}^n(2k-1)(g-1)=(g-1)(n^2-1)\) from the differentials plus the \(\mathcal{T}(\Sigma)\)-factor \(6g-6\)... the clean count is \(\dim\mathrm{Hit}_n(\Sigma)=(2g-2)(n^2-1)\).
Theorem (Labourie 2006). Every Hitchin representation \(\rho\in\mathrm{Hit}_n(\Sigma)\) is \(B\)-Anosov (see Anosov representations) and hence discrete, faithful, and a quasi-isometric embedding. The boundary map \(\xi_\rho:\partial\pi_1\Sigma\to\mathbb{RP}^{n-1}\) is a Frenet curve. Conversely (Labourie, Guichard–Wienhard), a representation into \(\mathrm{PSL}_n(\mathbb{R})\) with a Frenet boundary curve is Hitchin.
Maximal representations¶
Let \(G\) be of Hermitian type: \(G/K\) carries a \(G\)-invariant Kähler form \(\omega_{G/K}\). The Toledo invariant of \(\rho:\pi_1\Sigma\to G\) is
(independent of the choice of \(f\)). The generalized Milnor–Wood inequality bounds it:
Theorem (Burger–Iozzi–Wienhard). For tube-type Hermitian \(G\) of rank \(r=\mathrm{rank}_{\mathbb{R}}(G)\), $$ |\mathrm{T}(\rho)|\le|\chi(\Sigma)|\,r. $$ The maximal components \(\{|\mathrm{T}(\rho)|=|\chi(\Sigma)|\,r\}\) are unions of connected components of \(\mathrm{Rep}(\pi_1\Sigma,G)\) and consist of discrete and faithful representations. Each maximal representation admits a transverse dynamics-preserving boundary map into the Shilov boundary \(\check\partial(G/K)\) and is partially \(P\)-Anosov (for the parabolic stabilizing the tube-type Hermitian symmetric space).
For \(G=\mathrm{Sp}(2n,\mathbb{R})\) the maximal components generalize the classical \(\mathrm{Sp}(2,\mathbb{R})\cong\mathrm{SL}_2(\mathbb{R})\) story (the Toledo invariant is the Euler number for \(n=1\)). For \(G=\mathrm{SO}_0(2,n)\) the maximal components parametrize maximal immersed hyperbolic surfaces in \(\mathbb{H}^n\).
Non-abelian Hodge theory (the bridge)¶
The Higgs-bundle parametrization of Hitchin components is a consequence of the non-abelian Hodge correspondence (Corlette, Donaldson, Hitchin, Simpson):
| Theorem (reference) | Statement |
|---|---|
| Corlette; Donaldson | For a reductive \(\rho:\pi_1\Sigma\to G\) there is a \(\rho\)-equivariant harmonic map \(f:\widetilde\Sigma\to G/K\); existence and uniqueness up to flat sections. |
| Hitchin–Simpson | Reductive \(G\)-representations of \(\pi_1\Sigma\) \(\leftrightarrow\) polystable \(G\)-Higgs bundles on \(\Sigma\) (with appropriate real form / \(G\)-structure). |
| Hitchin fibration | The Higgs-bundle moduli \(\mathcal{M}_{\mathrm{Higgs}}(G)\to\bigoplus H^0(K^{\otimes d_i})\) (char. polynomial map) is an integrable system; the generic fiber is an abelian variety. |
| \(\mathbb{C}^*\)-action | The Higgs-bundle moduli carries a \(\mathbb{C}^*\)-action \((E,\varphi)\mapsto(E,\lambda\varphi)\); this is the Hitchin fibration, not an action on the character variety. |
The dictionary is: a Hitchin representation \(\rho\in\mathrm{Hit}_n(\Sigma)\) corresponds, via non-abelian Hodge, to a polystable \(\mathrm{PSL}_n(\mathbb{R})\)-Higgs bundle \((E,\varphi)\) with \(E\) a rank-\(n\) vector bundle and \(\varphi\in H^0(\mathrm{End}(E)\otimes K)\) a Higgs field; the Hitchin component is the component of the Higgs moduli on which the associated representation is real and Hitchin. The Hitchin section picks the Higgs bundle with \(E=K^{(n-1)/2}\oplus\cdots\oplus K^{-(n-1)/2}\) (split, degree-zero in the \(\mathrm{PSL}_n\) sense) and \(\varphi\) determined by \((q_2,\dots,q_n)\).
Hitchin vs maximal: comparison table¶
| Aspect | Hitchin components | Maximal components |
|---|---|---|
| Lie group \(G\) | split real (e.g. \(\mathrm{PSL}_n(\mathbb{R})\), \(\mathrm{SO}(n,n+1)\), split \(\mathrm{SO}(p,q)\), split exceptional) | Hermitian type (e.g. \(\mathrm{Sp}(2n,\mathbb{R})\), \(\mathrm{SO}_0(2,n)\), \(\mathrm{SU}(p,n-p)\), \(\mathrm{SO}^*(2n)\), exceptional Hermitian) |
| Defining input | principal \(\mathrm{SL}_2(\mathbb{R})\hookrightarrow G\); component of Fuchsian locus | Toledo invariant \(\mathrm{T}(\rho)\) at its bound $ |
| Invariant | (none / holomorphic differentials via Higgs) | Toledo invariant / Kähler class |
| Parametrization | Hitchin section: \((q_2,\dots,q_n)\), \(q_k\in H^0(K^{\otimes k})\) | via Higgs bundles with Toledo bound; Cayley correspondence (tube-type) |
| Anosov flavor | \(B\)-Anosov (Labourie) | partially \(P\)-Anosov (Burger–Iozzi–Wienhard; Guichard–Wienhard) |
| Boundary map | Frenet curve \(\partial\pi_1\Sigma\to G/B\) | transverse map to Shilov boundary \(\check\partial(G/K)\) |
| Dimension | \((2g-2)\dim G\) | \((2g-2)\dim G\) (smooth components) |
| Reference | Hitchin 1987/1992; Labourie 2006 | Burger–Iozzi–Wienhard 2010; for \(\mathrm{Sp}(2n,\mathbb{R})\) Goldman, Gothen, etc. |
See it¶
Toledo invariant for a maximal representation into \(\mathrm{Sp}(2n,\mathbb{R})\). Let \(\rho:\pi_1\Sigma\to\mathrm{Sp}(2n,\mathbb{R})\) be a maximal representation. Under the Higgs-bundle dictionary, \(\rho\) corresponds to a polystable \(\mathrm{Sp}(2n,\mathbb{R})\)-Higgs bundle, equivalently a rank-\(2n\) symplectic vector bundle \(E\) with Higgs field \(\varphi:E\to E\otimes K\) skew-symmetric. The Toledo invariant is the degree of the maximal isotropic subbundle: write \(E=L\oplus L^\vee\oplus W\) with \(L\) a maximal isotropic of degree \(\deg L\); then \(\mathrm{T}(\rho)=2\deg L\) (up to normalization), and the Milnor–Wood bound \(|\mathrm{T}|\le|\chi|\,n\) becomes \(|\deg L|\le(g-1)n\). Maximality forces \(L\) to be stable of extremal degree, and the Higgs field degenerates into a specific shape (the Cayley correspondence identifies maximal \(\mathrm{Sp}(2n,\mathbb{R})\)-Higgs bundles with \(\mathrm{SO}(n,n)\)-twisted Higgs bundles, the "Cayley partner"). Concretely for \(n=1\) this recovers \(\mathrm{Sp}(2,\mathbb{R})\cong\mathrm{SL}_2(\mathbb{R})\) and \(\mathrm{T}=\mathrm{e}(\rho)=2\deg L\in\{\pm(2g-2)\}\) — the classical Euler number; the maximal components are the two Teichmüller spaces. For \(n\ge 2\) the maximal components are higher Teichmüller spaces that are not Hitchin components (different Lie group, different invariant), and the boundary map lands in the Shilov boundary \(\check\partial(\mathbb{H}^n_{\mathrm{sym}})=S^{n-1}\times S^{n-1}/\mathbb{Z}_2\) rather than in a projective flag. This single computation — "the Toledo invariant is twice the degree of the maximal isotropic, and its bound is the maximal-degree condition" — is the operational content of the maximal paradigm.
Check¶
Explain why the \(\mathbb{C}^*\)-action on the Higgs-bundle moduli does not descend to an action on the character variety, and why this means the Hitchin fibration is not a fibration of representations.
The non-abelian Hodge correspondence identifies the polystable Higgs-bundle moduli \(\mathcal{M}_{\mathrm{Higgs}}(G)\) with the character variety \(\mathrm{Rep}(\pi_1\Sigma,G)\) (reductive reps) as real analytic spaces, but the identification depends on the choice of harmonic metric (the equivariant harmonic map). The \(\mathbb{C}^*\)-action \((E,\varphi)\mapsto(E,\lambda\varphi)\) changes the Higgs field, hence the harmonic metric, and the associated representation changes by a non-trivial deformation in the character variety — but the map \(\lambda\mapsto\rho_\lambda\) is not a \(\mathbb{C}^*\)-action on \(\mathrm{Rep}\) (there is no \(\mathbb{C}^*\)-action on a real character variety; the variety is real-analytic, not complex-algebraic in this sense). The Hitchin fibration \(\mathcal{M}_{\mathrm{Higgs}}\to\bigoplus H^0(K^{\otimes d_i})\) is therefore a fibration of Higgs bundles, and the generic fiber is an abelian variety (the Prym variety of the spectral cover), not a family of representations up to some action. Conflating the two is the standard error: the Hitchin component as a submanifold of \(\mathrm{Rep}(\pi_1\Sigma,G)\) is diffeomorphic (via non-abelian Hodge) to a component of the Higgs moduli, but the Hitchin fibration (the integrable system structure) lives on the Higgs side and does not transfer to a fibration of the character variety by complex orbits.
For \(G=\mathrm{Sp}(2n,\mathbb{R})\) (which is both split real and Hermitian), are the Hitchin and maximal components the same? Justify by comparing the defining conditions and the Anosov flavor.
No, for \(n\ge 2\) they are genuinely different components. The Hitchin component \(\mathrm{Hit}(\Sigma,\mathrm{Sp}(2n,\mathbb{R}))\) is the component of the Fuchsian locus under the principal \(\mathrm{SL}_2\hookrightarrow\mathrm{Sp}(2n,\mathbb{R})\); it is \(B\)-Anosov (Labourie). The maximal components \(\{|\mathrm{T}|=|\chi|\,n\}\) are partially Anosov but not \(B\)-Anosov in general (only the tube-type directions contract). The two components have the same dimension \((2g-2)\dim\mathrm{Sp}(2n,\mathbb{R})\) but are distinct submanifolds of \(\mathrm{Rep}(\pi_1\Sigma,\mathrm{Sp}(2n,\mathbb{R}))\), distinguished by the Toledo invariant (constant on maximal components, varying on the Hitchin component — except that the Hitchin component also has maximal Toledo invariant in some conventions, in which case it is one of the maximal components but not the whole maximal locus). The clean statement (Burger–Iozzi–Wienhard; Gothen): the Hitchin component is contained in a maximal component for \(\mathrm{Sp}(2n,\mathbb{R})\) only for \(n=1\); for \(n\ge 2\) the maximal locus is a union of components and the Hitchin component is one of several. The two paradigms give different higher Teichmüller spaces in the same Lie group.
Pitfalls¶
- Non-abelian Hodge identifies representations with Higgs bundles, but the \(\mathbb{C}^*\)-action / Hitchin fibration live only on the Higgs side: do not transfer the integrable-system structure to the character variety; the correspondence is real-analytic, not \(\mathbb{C}^*\)-equivariant.
- Hitchin requires split real \(G\): the principal \(\mathrm{SL}_2\) and the irreducible \(\mathrm{PSL}_2\hookrightarrow\mathrm{PSL}_n(\mathbb{R})\) exist only for the split real form; for \(\mathrm{SU}(p,n-p)\), \(\mathrm{SO}^*(2n)\), or other non-split real forms there is no Hitchin component in the same sense (one uses maximal instead).
- Maximal \(\neq\) Hitchin: for \(\mathrm{Sp}(2n,\mathbb{R})\) (\(n\ge 2\)) both paradigms exist but give different components; the Hitchin component is contained in the maximal locus only in low rank (\(n=1\)). Conflating the two is the standard error of the field.
- Toledo integrality: \(\mathrm{T}(\rho)\in\frac{1}{d_G}\mathbb{Z}\) where \(d_G\) depends on \(G\) (e.g. for \(\mathrm{Sp}(2n,\mathbb{R})\), \(d_G=1\); for \(\mathrm{SO}_0(2,n)\) the lattice is finer). The Milnor–Wood bound must be checked in the correct normalization.
- Tube-type hypothesis in Burger–Iozzi–Wienhard: the sharp discreteness theorem for maximal reps is proved for tube-type \(G\); for non-tube-type Hermitian \(G\) one combines tube-type reduction with partial Anosov arguments, and the statement is subtler.
Refs¶
- Hitchin, The self-dual Yang–Mills equations over Riemann surfaces (1987); Lie groups and Teichmüller space (1992)
- Labourie, Anosov flows, surface groups and curves in projective space (Invent. 2006)
- Burger–Iozzi–Wienhard, Surface group representations with maximal Toledo invariant (Annals 2010)
- Corlette, Flat \(G\)-bundles with canonical connections; Donaldson, Twisted harmonic maps
- Simpson, Higgs bundles and local systems; Gothen, The topology of the moduli space of \(\mathrm{Sp}(2n,\mathbb{R})\)-Higgs bundles
Next¶
Anosov Representations · Frontiers · Higher Teichmüller index · Foundations · Complex geometry / Kähler · Representation theory